Microsoft presents MAI-Image-1, its first own AI model to generate images

Microsoft presents MAI-Image-1, its first own AI model to generate images

When ChatGPT broke out at the end of 2022, Google began to hesitate -that bard that preceded Gemini– and Microsoft He reacted immediately. reached an agreement with OpenAI to use your language models GPT and began to integrate them into all its products. But Microsoft’s intention is not to always depend on the technology of others, and this year it has begun to take steps to stop doing it from OpenAI.

To the family of small language models Phi His first two LLMs (large language models) followed in August: the speech generator MAI-Voice-1 and the chatbot MAI-1-preview. On the other hand, Microsoft has recently begun to use Anthropic for some functions Microsoft 365. And these alternatives are now joined MAI-Image-1a text-based image generator designed and developed entirely within the company. The technology giant has described this new LLM, which will compete with gpt-image-1 of OpenAI and models Gemini/Image from Google, like ‘the next step on our path’.

According to the company, the new MAI-Image-1 model has been trained with the aim of offering real value for creators. The team focused especially on prevent the model from producing repetitive results, with generic or excessively stylized stylesthe so-called ‘AI look’ that can be seen in most synthetic generations.

Microsoft claims to have achieved this through rigorous data selection and precise task-oriented evaluation. Furthermore, ‘MAI-Image-1 stands out for generate photorealistic images, such as lighting (e.g. bounced light or reflections), landscapes and more’explains the company.

Microsoft also ensures that this new model is considerably faster than other large sizes available on the market. According to the company, users will be able to generate the images they want more quickly and perform iterations more agilely.

In Image Arena of LMArenaan open community evaluation platform for AI models, MAI-Image-1 debuted at number 9. Although it may not seem like a spectacular result, it represents an important achievement for a first-generation model developed entirely by Microsoft. In the classification, ByteDance, OpenAI, Google and Tencent They occupy superior positions. The CEO of Microsoft AI, Mustafa Suleymanpointed out in X that they will continue to perfect the model to climb positions in the ranking.

When will it be available? Microsoft assures that bringing you to Copilot and Bing Image Creator ‘very soon’.